Scale-Invariant Decision-Making

• Scale-Invariant Decision-Making is defined by aggregation rules that remain consistent across varying group sizes and contexts.
• Empirical studies reveal power law scaling (α, β, γ ≈ 1/3) in local elections, indicating robust turnout and representation metrics.
• Triadic deliberation achieves near-optimal consensus with O(n log² n) interactions, while spatial models exhibit fractal bifurcation patterns ensuring bounded decision processes.

Scale-invariant principles of decision-making refer to structural and dynamical laws governing how groups, agents, or systems aggregate decisions in ways that are preserved across changes in scale—whether by number of agents, the size of the choice set, or spatial extent. These principles have been elucidated in diverse contexts including electoral systems, deliberative protocols, and animal navigation. Common threads include fractal hierarchies, emergent invariants in collective statistics, and recursive group-structuring rules that yield robust group-level outcomes independently of system size.

1. Empirical Scaling Laws in Social Decision-Making

Empirical evidence for scale invariance in decision-making emerges most clearly in group-outcome settings where the decision consequences reference only the specific group undertaking the choice. Large-scale studies of local elections reveal that both participation rates and institutional properties obey nontrivial scaling with group size. For each municipality with $N$ members, the logarithmic turnout rate $\tau = \ln(N_{+}/N_{-})$, with $N_{+}$ as votes cast and $N_{-}$ as abstentions, exhibits a mean and variance obeying universal power laws: $$\langle\tau\rangle(N) \approx C - \alpha\ln N, \qquad \alpha \approx 1/3,$$

$$\sigma^2_\tau(N) \approx D/N^{\beta}, \qquad \beta \approx 1/3,$$

across multiple countries and elections, for strictly local contexts. Similarly, the total number of democratic representatives $n_r$ in a decision-making body scales sublinearly: $n_r \sim E N^\gamma, \qquad \gamma \approx 1/3,$ independent of polity scale or geography. No such invariance holds for national or supra-group elections when plotted against local unit size. This statistical robustness is indicative of an underlying structural universality (Borghesi et al., 2013).

2. Fractal-Hierarchy Interpretation and Theoretical Frameworks

The coincident scaling exponents ($\alpha \approx \beta \approx \gamma \approx 1/3$) motivate a fractal-hierarchical interpretation of group organization. Any constituency of size $N$ is modeled as composed of $n_g(N) \approx N^{\delta}$ ($\delta \approx 1/3$) subgroups at each hierarchical level. This is supported both by variance decomposition—where the variance in aggregate turnout attributable to subgroup structure decays as $N^{-\delta}$—and by analysis of mean turnout slopes with respect to $\ln N$, which similarly reflect subdivision at each scale. Phenomenological models assign each agent a “conviction” $C_i = H(i) - d(i) + F_{\rm nat}$: $H(i)$ are group-level inherited biases, $d(i)$ the depth in the hierarchy, and $F_{\rm nat}$ a group-wide effect. Logistic voting choices and hierarchical structure together yield the empirically observed turnout decay and variance across scales (Borghesi et al., 2013).

A plausible implication is that group-decision systems exhibiting this scale invariance reflect recursively nested subgroups with minimal “depth cost” and persistent noise-splitting at each level, producing robust critical exponents regardless of constituency scale.

3. Scale-Invariant Deliberative Consensus: Role of Small Group Protocols

Scale-invariant consensus formation extends to the analysis of deliberative decision-making protocols. Representing participant opinions as nodes in a median graph $G=(V,E)$, the task is to recover the generalized median $x^* = \arg\min_{x \in V}\sum_{i=1}^n d(x, x_i)$—a rigorous proxy for the “wisdom of the crowd.” A class of protocols based on repeated triadic (three-person) majority-rule interactions implements a token-passing process:

• In each round, select three tokens (participants) at random.
• These participants deliberate (majority-rule) to a Condorcet winner, or the unique median on $G$.
• Tokens are transferred to the local winner. The process repeats until all tokens coalesce at a single node.

The process enjoys provably scale-invariant convergence:

• $O(n \log^2 n)$ triadic interactions suffice for $(1+O(\sqrt{\log n/n}))$-approximation of $x^*$, with per-participant effort scaling as $O(\log^2 n)$.
• The error vanishes as $n \to \infty$; thus, quality and scalability are preserved as group size increases (Goel et al., 2016).

Crucially, dyadic (pairwise) decision protocols fail to achieve this vanishing error, instead incurring a constant $(9/8)$-factor loss in worst-case configurations. The ability of triads to filter outliers and construct robust local medians is essential to scale invariance in global consensus.

4. Geometric Structure and Self-Similarity in Spatial Decision-Making

In spatial decision-making, such as animal navigation among discrete targets, scale invariance emerges in the geometry of trajectories and choice bifurcations. Dynamical models comprising groups of Ising spins (neural “activity bumps”) associated with spatial targets yield a consensus-driven velocity vector for the agent or collective: $\vec V = \bar v \sum_{i=1}^k n_i \hat p_i,$ where $n_i$ are mean active fractions and $\hat p_i$ unit vectors to each target. The underlying spin-system Hamiltonian encodes local excitation and long-range inhibition, and the mean-field reduction yields a set of algebraic equations for steady-state velocities and neural activations.

The solutions of these mean-field equations define “mean-field trajectories” in physical space, whose bifurcations are governed by analytically determined “bifurcation curves.” These curves demarcate regimes of spontaneous symmetry breaking in neural consensus, yielding:

• Self-similar (fractal) binary branching of trajectories when bifurcation curves intersect appropriately, with a geometric progression in successive split distances.
• Space-filling tree-like exploration when curves braid without intersection.

Scale invariance is thus encoded in the geometry: regardless of initial agent position, the set and structure of bifurcation curves fully determine the sequence of decisions and their scales (Gorbonos et al., 2023).

5. Impact of Metric Geometry: Non-Euclidean Embedding and Decision Boundedness

Biologically realistic modifications to the metric structure, in particular replacing the Euclidean dot-product with a nonlinear, “elliptic” angle mapping

$$\cos \theta_{ij} \longrightarrow \cos[\tilde\theta_{ij}], \quad \tilde\theta_{ij} = \pi \left( \frac{\theta_{ij}}{\pi} \right)^\nu, \ 0<\nu\leq1,$$

alter the interaction topology among targets. For $\nu<1$—consistent with empirical neural data—bifurcation curves spread and the number of possible bifurcations collapses, enforcing a bounded decision-tree depth. Thus, the process eliminates pathological infinite indecision and enables uniform, bounded exploration, sharply enforcing scale invariance at the level of neural computation. This suggests a functional evolutionary role for non-Euclidean spatial representations in facilitating efficient collective decision-making (Gorbonos et al., 2023).

6. Implications, Limitations, and Open Questions

The confluence of universal exponents across social, deliberative, and spatial settings indicates a deep structural basis for scale invariance in decision-making systems. The emergence of hierarchical recursive organization, robust mean and fluctuation scaling, and the necessity of specific group sizes (e.g., triads vs dyads) is consistent with both empirical and theoretical frameworks.

Nevertheless, several open questions persist. The origin of the fractal hierarchical subdivision in social systems remains unresolved—possibilities include cognitive constraints (e.g., logarithmic perception of group size), cultural/institutional rules, or selection for optimal assembly cost. Likewise, the mean-field spatial decision models are phenomenological; deriving the precise form and parameters of neural interaction from first principles remains an active area of research. Finally, the boundary conditions for the breakdown of scale invariance—especially in contexts where decisions have extra-group consequences—are not fully understood. These factors delineate ongoing directions in the investigation of scale-invariant decision-making (Borghesi et al., 2013, Goel et al., 2016, Gorbonos et al., 2023).

---

Scale-Invariant Setting	Principle/Structure	Key Exponent(s) / Property
Local-election turnout	Turnout and variance scale as $N^{-1/3}$	$\alpha \approx \beta \approx 1/3$
Institutional representation	Representatives grow as $N^{1/3}$	$\gamma \approx 1/3$
Triadic deliberation	$O(\log^2 n)$ per-agent rounds, error vanishes	$\varepsilon = O(\sqrt{(\log n)/n})$
Spatial bifurcation geometry	Mean-field bifurcations encode choice splits	Self-similarity, space filling